Multi-Agent Systems involving large numbers of agents are difficult to design and implement. The task can be made more manageable through the use of agent protocols. Agents protocols define conventions which simplify cooperation between agents. One method of defining agent protocols is the Electronic Institutions model discussed in previous talks. However, there are a number of potential drawbacks of this approach, relating to the static nature of the Institutions. In this talk I will outline a technique which overcomes this problem by allowing agent protocols to be specified in a flexible manner as the agents communicate.